spikeSlabGAM (version 1.0-0)

plotTerm: Plot the estimated effect of a model term.

Description

Plot the estimated effect of a model term.

Usage

plotTerm(label, m, cumulative=TRUE, aggregate=mean, quantiles=c(0.1,
    0.9), gridlength=40, contours=30, ggElems=list())

Arguments

label

(character) the label of the effect to be plotted.

m

a fitted spikeSlabGAM model

cumulative

Defaults to TRUE, in which case all lower order terms that are involved in label are cumulated and then plotted. (e.g, if label denotes a smoooth term, the sum of the linear and smooth effect is displayed, if label

aggregate

(function) which summary statistic to use for the posterior of the effect. Defaults to the mean.

quantiles

which quantiles to use for the borders of credible regions. Defaults to 10Set to NULL to omit plotting credible regions. Cannot deal with more than two quantiles.

gridlength

length of the (univariate) grids on which to evaluate the posterior.

contours

use how many contour lines for the joint effect of two numerical covariates. Defaults to 30.

ggElems

a list of plot elements to give to ggplot. Use this to supply custom themes, colour schemes etc. Defaults to an empty list, which yields the standard settings.

Value

• an object of class ggplot. Use print or wrap the call to plotTerm in parentheses to directly render on a graphic device.

Details

Plots the estimated linear predictor for a model term, i.e a main effect or a (2- or 3-way) interaction. Plots for the joint effect of two numerical covariates show an overlay if quantiles were specified: Regions where the pointwise credible intervals do not contain zero are plotted in muted red ($>0$) and blue ($< 0$), overlaid by coloured contour lines that show the aggregate values. Contour lines are shown only inside the convex hull of the original observations. Plots for link{srf}:link{lin} terms show the spatially varying coefficient, i.e. the contour lines represent the change in the linear predictor when the lin-covariate increases by 1 standard deviation. For this reason, a cumulative plot makes no sense and the routine will set cumulative=FALSE with a warning. Limitations: Plots for 4-way (or higher) interactions are not implemented. Requesting such a plot will return the NULL object with a warning. Plots for mrf just treat the grouping variable as a conventional factor i.e. will not incorporate any neighborhood or map information.

Examples

#see help for spikeSlabGAM}